Data-driven approach for potential iron-based half-Heusler thermoelectrics with chemical bonding characteristics.

Chemical bonding influences various physical properties and holds promise for guiding the discovery of high-performance materials. In XYZ half-Heusler (HH) compounds, complex interactions between constituent atoms, featuring both covalent and ionic characters, are supposed to affect mechanical and thermoelectric behavior. We use a data-driven approach based on first-principles calculations to identify promising HH compounds and explore the correlations between bonding-related features and functional properties. Our analysis suggests that enhanced Y─Z bonding correlates with greater bulk modulus (B), larger Grüneisen parameter, and enhanced power factor. Tungsten-iron-lead stands out with excellent B = 162.4 gigapascals and a low lattice thermal conductivity of ~7.7 watts per meter per kelvin, leading to a figure of merit of ~0.52 (at 526 kelvin) without any nanostructuring, surpassing that of vanadium-iron-antimony by ~80.8%. These findings highlight how chemical bonding characteristics, interpreted from electronic structure and orbital-resolved bonding analysis, provide in-depth insights into the structure-to-property correlation to accelerate the screening out of potential thermoelectrics.